
Universitetslektor i kognitionsvetenskap med inriktning mot lingvistik @umeauniversity. Talforskare, kognitionspsykolog, skribent, debattör, och lite till.
Umeå / Stockholm
Joined August 2017
- Tweets1.3K
- Following497
- Followers211
- Likes20.4K
Intressant om bias hos språkmodeller.
Researchers proved every major LLM is secretly biased against men.
And they finally figured out why.
They evaluated leading LLMs across a wide range of real-world conflict scenarios, from workplace disputes to personal and ethical dilemmas.
They wrote strictly gender-neutral misconduct templates and turned them into paired first-person prompts.
Same mistake. Same context. Same severity.
Only thing they changed was whether the person involved was a man or a woman.
Then they ran those pairs across 10 representative LLMs and tracked five signals in the responses: punishment, empathy, escalation, instruction, and blame.
And the results exposed a consistent, asymmetrical machine morality.
When the actor was a man, the AI turned harsher. It reached for punitive language, escalated the conflict, lectured the user, and assigned blame directly.
When the actor was a woman, for the exact same misconduct, the AI shifted gears completely.
The tone became therapeutic. Contextualizing. Empathy-oriented.
Then the researchers tested whether smarter models fixed the problem. They didn't.
They checked across different model families. They scaled the models up. They turned on explicit thinking-style reasoning.
The double standard survived all of it. Even when the AI was given time to reason step by step, it still judged men more harshly than women for identical behavior.
Nobody explicitly coded these rules.
It is baked into the open-ended framing of the models millions of people now use as neutral referees for their relationships and conflicts. Same offense, different verdict, hidden behind a helpful tone.
18
Axel Ekström retweeted
Social scientists (especially researchers at university) are known to be more politically left than the general population. But does that tendency also appear in the research they produce? A recent article in the journal @Theory_Society says that the answer is yes. The author used AI to rate almost 600,000 abstracts for articles published between 1960 and 2024 in 11 disciplines in the social sciences on a scale of 0 (far right) to 10 (far left). The study found that:
➡️Only 30.1% of abstracts contained information that is politically relevant in the mid 2020s. Most social science research isn't ideological.
➡️However, when social science does relate to political issues, it's almost always from a leftist perspective. Of the politically relevant abstracts, approximately 90% support views that are left of center.
➡️Gender studies was the most left-leaning discipline (average abstract = 7.6) and economics was the most centrist (average = 5.7; see 2nd image).
➡️In every discipline, the majority of politically relevant abstracts were rated between 6 and 8, which corresponds to the range of views found between Joe Biden (6) and Alexandria Ocasio-Cortez (8).
➡️Economics was the only discipline that had a nontrivial number of articles that had right of center perspectives. 16% of its politically relevant abstracts were rated at 4 (equivalent to Tom Cotton) or lower.
➡️Every discipline's output has become more politically left over time, but this is more true for disciplines which are more theoretical and less connected with policy (e.g., sociology, psychology, ethnic studies). For disciplines that study policy (e.g., criminology, public administration, political science), the output grew more centrist in the 1970s and 1980s before becoming more left-leaning starting in the 1990s (see 3rd image).
➡️The more left-leaning a discipline's output was, the less ideological diversity can be found within it.
➡️The leftward trend of research abstracts accelerated in all fields, starting about 2010. This trend has been much stronger for sociocultural issues than for economic ones (see 4th image).
This study doesn't prove an ideological bias, but the results do show that social science research is dominated by politically leftist output. The author suggests that fields more sensitive to policy issues and political influences may be more responsive to broad political and social trends (allowing them to become more centrist in the 1970s and 1980s) but that theoretical fields do not have this tempering influence. As time passes, certain disciplines become more ideologically identified, and self-selection into the field attracts new left-of-center researchers and homogenizes the field over time.
Read the full study (with no paywall) at doi.org/10.1007/s11186-026-0…
Det händer mycket intressant i intelligensforskningens värld.
New from David's Reich's lab: "In the past ten millennia [in Europe], we find that many hundreds of alleles have been affected by strong directional selection," including those influencing human intelligence.
nature.com/articles/s41586-0…
1
136
USAs största organisation för akademisk psykologi har valt att upphöja en postmodern ordmånglare till ledarskapsposition. Det är såhär man övertygar folk att forskare inte tillför samhället något.
Skäms, de som urholkar sina vetenskaper.
Candice Hargons is the incoming president of the American Psychological Association, the main professional organization of US psychologists. Here are some of her publications
1
165
Axel Ekström retweeted
Replying to @MagisterDave @TireddadtalesSE
Det är så.
För att kunna fokusera på innehållet i en mening eller paragraf krävs att byggstenarna i form av ord 'kommer av sig själva', dvs. utan ansträngning (= läsandet/skrivandet av dem är automatiserat).
Den som behöver ägna tankekraft och tid åt stavningen fastnar i detta och har ingen/begränsad bearbetningskapacitet kvar till innehållet.
Vid skrivande gäller detta även den som använder stavningskontroll eftersom detta moment bryter av och kräver tid (jämfört med att 'stava rätt' då 'det sitter i ryggmärgen').
På liknande sätt vid läsning: Ord vars stavning eleven inte har automatiserat kräver arbetsminneskapacitet - och risken är att ingen kapacitet finns kvar för att bearbeta och förstå innehållet i meningarna/paragrafen ifråga.
Så ja - vill vi att elever ska läsa och förstå det de läser behöver de (bland annat) få incitament och stöd - och tid - för att bli skickliga på att stava.
Betydelsefull analysis av veckans PISA-släpp. Stämmer detta har många hoppat på grumliga framställningar.
Did student ability plummet on the PISA examination this year?
No, they did not experience any meaningful change. The large observed change is due to two factors:
1. Measurement issues (which could not be fully corrected due to a format change)
2. Sampling changes
7
693
Axel Ekström retweeted
Välskrivet av Tarja Alatalo men också beklagligt att texter som denna behöver skrivas.
vilarare.se/nyheter/vi-larar…
Axel Ekström retweeted
Tämä suomalaisella datalla tehty kesällä 2026 taloustieteen huippulehdessä (AER) julkaistu tutkimus on saanut todella vähän mediahuomiota Suomessa.
Tulosten mukaan Suomen opettajakoulutuksen mieskiintiön poisto vuonna 1989 huononsi sekä poikien että tyttöjen tuloksia. (1/2)
Axel Ekström retweeted
Om svenska ingångslöner för lärare i åk 7–9 ska ligga i topp i OECD krävs en fördubbling: 70–80 000 kr/mån. @liberalerna´s +10 000 bara en början. Övriga partier blundar för att Sverige ligger först på plats 11. Gapet till Tyskland och Luxemburg kvarstår
ekonomifakta.se/sakomraden/u…
Mycket bra från en klartänkt elev.
Rimliga synpunkter från eleven Hanna.
vilarare.se/nyheter/vi-larar…
Pinsamt hanterat av ett av världens äldsta och mest anrika universitet.
Ultimately, Cambridge University will be a massive loser in this affair, deservedly so. It is a necessary condition of the place’s continuing glamour and prestige that it enforces rigorously objective academic standards. To be shown not to care about this, in such a public way, is an act of catastrophic institutional self-harm.
Read @docstockk on Jason Arday: Cambridge’s DEI darling ⬇️
unherd.com/2026/07/jason-ard…
2
335
Axel Ekström retweeted
My scholarly article "Why There Are Exactly Two Sexes" has been accessed over 50K times, placing it in the top 8 of all articles in the journal's history.
Activists refuse to cite it or acknowledge it exists.
I'm still waiting for someone to publish a scholarly scientific takedown of it. You'd think this would be a top priority for those asserting "sex is a spectrum" who believe my work is bigoted, pseudoscientific, and actively harming a vulnerable community.
Until then, you can read it and arm yourself with a thorough understanding of the 5 major frameworks that activists and activist scientists use to undermine the binary nature of the sexes and why they all fail.
🔗link.springer.com/article/10…
Axel Ekström retweeted
“Consistent with the literature, the gender gap in teacher grading is against boys.”
Axel Ekström retweeted
While general intelligence is known to predict life success, a new study reveals that specific mental skills—like technical knowledge or math ability—play a surprisingly large role in determining a person’s future income, education, and career path. dlvr.it/TTfTJV
Axel Ekström retweeted
Replying to @stephaniemlee
Good God. She genuinely thinks Fighting the Power is the essence of serious thinking, has never been questioned about it even this far into her career, and is incapable of even understanding what Lee's questions are getting at. She's not deflecting. She really has no idea how ridiculous she sounds.
I became an academic out of being curious about things. Gradually realizing since the 1990s that this kind of person is ordinary in the humanities and social sciences, is routinely given awards and high administrative positions, and is beyond the reach of rational discussion has been one of the most disappointing experiences of my life.
Replik ute idag i Vi Lärare:
vilarare.se/nyheter/vi-larar…
Axel Ekström retweeted
This reminds me of a quote by evolutionary biologist Eörs Szathmary: "linguists would rather share each other's toothbrush than each other's terminology."


































